Attempts have been made to produce epitaxial layers from materials such as GaN, GaAs, InP, GaAlAs, InGaAs, AlN, AlGaN, and even SiGe. It is known that relatively thick layers of these materials, for example, layers having a thickness greater than 1 or 2 μm, are required for good crystalline qualities. The advantages of an epitaxial layer not being stressed (or only slightly stressed), and having a low defect density, i.e., having a dislocation density of less than 106/cm2 are also known.
Techniques such as MOCVD have been utilized to obtain epitaxial growth of thick layers of GaN, for example, a layer having a thickness greater than 12 μm on a substrate for epitaxy. The epitaxial growth of such GaN layers has been essentially on the bulk substrates of sapphire, SiC, or Si materials. These three substrates are the most frequently used since they are the most readily available substrates, although a few tests have been carried out on substrates such as ZnO or LiGaO2.
Currently, epitaxial GaN layers homogeneously deposited on a substrate surface have a dislocation density in the range 108 and 1010/cm2 regardless of the nucleation surface used. Additionally, the stresses in thick GaN layers obtained by MOCVD (growth temperature 1000-1100° C.) clearly depend on the coefficient of thermal expansion of the epitaxial substrate, which determines the stresses of thermoelastic origin that are imposed on the system.
For example, GaN layers produced on sapphire are in compression, while those obtained on SiC are under slight tension, and those on silicon are under high tension. Tension stresses produce a strong tendency for cracks to form in the epitaxial film, thereby destroying it. The layers subjected to the compression stresses, however, are also problematic.
These problems are particularly true for epitaxial growth on silicon substrates. For silicon epitaxy support, the limit beyond which cracks appear is about 1 μm to 2 μm, which is a limiting factor in producing thick, good quality epitaxial layers.
Growth tests on SOI (silicon on insulator) substrates have shown that the use of that type of substrate can reduce the crystal defect density in the epitaxially grown layer because of the compliant nature of the very thin film of silicon present on the oxide. However, that system is limited in its capacity to absorb stresses, in particular for thick GaN layers (like silicon, at best a film thickness of 1 μm to 2 μm).
It appears that the crystal quality improves by growth on substrates having motifs. The dislocation densities obtained are of the order of 106/cm2. Epitaxial lateral overgrowth (ELO) techniques exist, along with techniques known as pendeoepitaxy (PE), lateral overgrowth from trenches (LOFT), and cantilever epitaxy (CE). All of these techniques are based on lateral overgrowth and coalescence of the epitaxially grown layer to ultimately form a continuous film. The continuous films obtained have precise zones with improved crystal quality (the epitaxial lateral overgrowth (ELOG) technique), or have a homogeneous film of crystal quality (LOFT technique). Those solutions have been demonstrated for sapphire, SiC, and Si (111).
Although these solutions improve the crystal quality of the epitaxially grown film, they cannot effectively solve the problem of stress in the epitaxially grown films. Thus, a need exists for a substrate or a support that can absorb high levels of stress during crystal growth, and in particular during thick epitaxial growth of a material. In particular there is a need for a support that absorb stresses when the coefficient of thermal expansion of the epitaxial growth material is different from that of the substrate. The present invention now satisfies that need